A client's operations dashboard had a strong launch — daily logins, positive feedback in the first fortnight — and then usage dropped to almost nothing by week four. Nothing about the dashboard had changed.
The pattern that gave it away
Usage didn't decline gradually, it stepped down sharply once the novelty period ended, which is usually a sign the dashboard was answering a question people were curious about rather than one they needed answered regularly.
The question we should have asked earlier
Not "what data do you want to see" but "what decision does this change, and how often do you make that decision". The client's team made pricing and staffing calls weekly, not daily, and had built a habit of checking a spreadsheet a manager maintained by hand — which the dashboard had duplicated rather than replaced.
What we found once we looked properly
- The dashboard showed everything, decided nothing. Numbers with no threshold, no comparison, no "this is unusual" marker, so reading it took judgement the viewer had to supply themselves each time.
- The manual spreadsheet had informal annotations — a note next to a bad week explaining why — that carried more value than the chart it sat next to.
- Nobody had removed the old process, so the dashboard was extra work rather than replacement work for the first month.
What we rebuilt
We added explicit thresholds and comparisons — this week versus the trailing four-week average, flagged when a metric moved past an agreed band — and gave the team a way to annotate a data point, the one thing the old spreadsheet did that the dashboard didn't.
A dashboard that requires the viewer to already know what normal looks like is a display, not a tool.
The follow-up we now do
We schedule a usage check at week six on every dashboard product, specifically looking for the step-down pattern rather than a gradual decline, because the two have different causes. A gradual decline is usually data quality. A step-down is usually that the tool never replaced the habit it was meant to replace.